Embed Size (px)
Citation preview
This article was downloaded by: [Ams/Girona*barri Lib]On: 15 October 2014, At: 03:27Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954 Registeredoffice: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK
Journal of Prevention & Intervention inthe CommunityPublication details, including instructions for authors andsubscription information:http://www.tandfonline.com/loi/wpic20
Creating a Geospatial Database ofRisks and Resources to Explore UrbanAdolescent Substance UseMichael Mason a , Ivan Cheung b & Leslie Walker ca Villanova University , Villanova, Pennsylvania, USAb Insurance Institute for Highway Safety , Arlington, Virginia, USAc Children's Hospital and Regional Medical Center , Seattle,Washington, USAPublished online: 15 Jan 2009.
To cite this article: Michael Mason , Ivan Cheung & Leslie Walker (2009) Creating a GeospatialDatabase of Risks and Resources to Explore Urban Adolescent Substance Use, Journal of Prevention &Intervention in the Community, 37:1, 21-34, DOI: 10.1080/10852350802498391
To link to this article: http://dx.doi.org/10.1080/10852350802498391
PLEASE SCROLL DOWN FOR ARTICLE
Taylor & Francis makes every effort to ensure the accuracy of all the information (the“Content”) contained in the publications on our platform. However, Taylor & Francis,our agents, and our licensors make no representations or warranties whatsoever as tothe accuracy, completeness, or suitability for any purpose of the Content. Any opinionsand views expressed in this publication are the opinions and views of the authors,and are not the views of or endorsed by Taylor & Francis. The accuracy of the Contentshould not be relied upon and should be independently verified with primary sourcesof information. Taylor and Francis shall not be liable for any losses, actions, claims,proceedings, demands, costs, expenses, damages, and other liabilities whatsoever orhowsoever caused arising directly or indirectly in connection with, in relation to or arisingout of the use of the Content.
This article may be used for research, teaching, and private study purposes. Anysubstantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions
Creating a Geospatial Database of Risks andResources to Explore Urban Adolescent
Substance Use
MICHAEL MASONVillanova University, Villanova, Pennsylvania, USA
IVAN CHEUNGInsurance Institute for Highway Safety, Arlington, Virginia, USA
LESLIE WALKERChildren’s Hospital and Regional Medical Center, Seattle, Washington, USA
This article illustrates the methodology of creating a comprehensivegeospatial database in order to systematically understand thesocial ecology of risk and protection for urban youth. The chal-lenges and future opportunities involved with this complex workwere reviewed, and specific examples were provided to guideresearchers. Data were collected from a Washington, DC adoles-cent substance abuse treatment sample to construct a geospatialdatabase to evaluate urban youths’ social environmental riskand resources. A geographic information systems (GIS) approachwas adopted to integrate a large array of variables at differentlevels of geography. For example, risk factors included proximityto crime hotspots, and other known potential establishments withnegative influence (such as liquor stores). We also used GIS toassess the subjects’ accessibility to protective resources such as pub-lic libraries, recreational, parks, and police stations. Unique to ourmethod was the collecting and mapping of each teen’s activitylocations (places they typically frequent). These data form ‘‘riskand protection exposure’’ estimates for each teen. Finally, we
This research was supported by funding from the Substance Abuse and Mental HealthService Administration, Center for Substance Abuse Treatment, grant no. TI15433. The contentof this article does not necessarily reflect the views or policies of the government.
Address correspondence to Michael Mason, Villanova University, Department ofEducation & Human Services, St. Augustine Center, 800 Lancaster Ave., Villanova, PA 19010,USA. E-mail: [email protected]
Journal of Prevention & Intervention in the Community, 37:21–34, 2009
Copyright # Taylor & Francis Group, LLC
ISSN: 1085-2352 print=1540-7330 online
DOI: 10.1080/10852350802498391
21
Dow
nloa
ded
by [
Am
s/G
iron
a*ba
rri L
ib]
at 0
3:27
15
Oct
ober
201
4
illustrated the specific methods for creating a dynamic geospatialdatabase for urban youth and present future analyticalapproaches and challenges with these type of data.
KEYWORDS GIS, geospatial database, urban adolescent, sub-stance use, risks, resources
A social ecological research approach is necessary to understand thedynamic and fluid nature of modern urban youth health behavior. Thismethod explores the connections between teens’ mental health, their socialnetworks, and the everyday settings in which their health behaviors areexpressed (Mason, Cheung, & Walker, 2004). The current study was anexample of a research strategy for the study of substance-using urban youththat incorporated geographical factors in order to systematically understandthe social ecology of urban youth. It is possible that by better understandingthe interactions among individual, social, and environmental factors, preven-tion program effectiveness may be improved for different groups of youth(Sussman, Ames, Dent, & Stacy, 2001). This detailed, multilevel type ofresearch has been limited in part by the complex nature of this task and inpart by the lack of in-depth methodological approaches that can addressthe nuances of the social ecology of adolescent substance use. Our researchapproach was predicated on the premise that place and space are sociallyconstructed, complex living constructs with multiple meanings, and not merepassive containers in which things are simply recorded (Kearns & Moon,2002). For example, adolescents’ social networks are constituted by space(physical world), place (socially constructed), which influence our interpre-tations of meaning, and our sense of self. Social networks do not occur in aspaceless, decontextualized vacuum; rather they are constituted by the envir-onment (space and place) in which they operate. Place helps alter the flow ofour social interactions through space. Therefore, there is a mutual depen-dency between self and place. Place depends on self, and self depends onplace (Sack, 1997).
Very little neighborhood-effects research has addressed the ecologicalstructure of urban teens’ daily routine activities and the associated temporalcontingencies, that is, the influence of time of day, day of week, and amountof time spent in locations (Crouter & Larson, 1998; Sampson, 2003; Larson,1998; Macintyre & Ellaway, 2003; Raduenbush, 2003).
How to meaningfully define and measure geographic-based data is apoint of contention in the literature. There is concern that census tract dataare too large to meaningfully explain environmental influences, and conver-sely, subjectively identified neighborhoods are highly malleable concepts fullof personal meanings, varying locations, and interpretations (Berkman &Clark, 2003; Duncan & Aber, 1997; Furstenberg & Hughes, 1997; Larson,
22 M. Mason et al.
Dow
nloa
ded
by [
Am
s/G
iron
a*ba
rri L
ib]
at 0
3:27
15
Oct
ober
201
4
1998). Our research with urban youth informs us that the various locations inwhich they spend their time are not delimited by traditional geographicalboundaries such as census tracts, block groups, or political wards (Masonet al., 2004). Therefore we constructed the geographical area of each subjectaccording to his or her activity locations that they frequent. These highly indi-vidualistic geographical data stand in contrast with more traditionally used,although politically and seemingly arbitrary defined geographic units, suchas census tracts or block groups.
METHODS
During the last four years, we have developed our unique social ecologicalmethodology using geographic information systems (GIS). GIS can be under-stood as a powerful set of tools that captures, manages, analyzes, and visua-lizes spatial data (Burrough, 1986; Clarke, 2003). It can also be viewed as aspecial kind of information system in which information is related by itsshared spatial identity (Star & Estes, 1990). In sum, we use GIS to answerthe questions, ‘‘What is where?’’ and ‘‘Why it is there?’’ We use GIS to detectand recognize spatial patterns and investigate the processes that shape them.To read more about the details of our methods along with an illustrative casestudy see Mason et al. (2004).
Sample
We applied our method to our current study with urban adolescents enrolledin a substance abuse treatment program. This project was for Washington,DC–area teens that applies Motivational Enhancement Therapy and Cogni-tive Behavioral Therapy for five sessions and follows them for 12 monthsposttreatment. It was funded by the Substance Abuse and Mental HealthServices Administration, Center for Substance Abuse Treatment. The samplecomprised 58 teens (35% African American, 42% White, 17% of Hispanic ori-gin, and 6% other international origins) whose mean age was 17 years andmostly (82%) male.
Measures
The geographic data were collected using the Ecological Interview (Masonet al., 2004), which produces a geographically specific listing of the teen’sdaily activity locations, as well as evaluative descriptions of various geogra-phical environments. It produces a listing of safe (‘‘safest place from harm,danger, or the likelihood of engaging in risky or dangerous activities’’), risky(‘‘the place where you are most likely to engage in risky or dangerous activ-ities, cause trouble, or do illegal activities’’), and important (‘‘having the
Creating a Geospatial Database 23
Dow
nloa
ded
by [
Am
s/G
iron
a*ba
rri L
ib]
at 0
3:27
15
Oct
ober
201
4
biggest impact on your life, most meaningful’’) places that the adolescentmost often frequents. Teenagers were also asked about which people in theirsocial network are typically at each setting they have listed.1 The EcologicalInterview is a brief, structured interview based on the qualitative methodologyknown as ‘‘Free Listing’’ (Weller & Romney, 1998). It is based on an Environ-mental Indicators of Health Outcomes measure (Cheadle, Wagner, Koepsell,Kristal, & Patrick, 1992). This interview probes teens to generate a weeklyaccount of their activities, their routines, and travel patterns for a typical week,collecting exact geographic information (e.g. street addresses, cross streets,stores, parks, and landmarks). These data are used to construct the respectiveactivity space around each activity location, and thereby create risk and protec-tion exposures estimates. These locations were referenced to a geographicalframe using the street center-line data obtained from the DC Office of theChief Technology Offer (OCTO) and we used the Maryland State Plane coordi-nates measured in meters. This geocoding process allowed us to referenceall spatial information to the same reference frame using a unified coordinatesystem.
RESULTS
Initial Descriptive Results
There were 283 locations (186 unique locations collected during the base-line interview, 62 during the 3 months follow-up, 29 during the 6 monthsfollow-up, and 6 during the 12 months follow-up) obtained from the 58 sub-jects during the various phases of the interviewing process. These 283 loca-tions were not necessary unique locations as a location can be repeatedlyentered into the database multiple times (e.g., home). Each location maythen be qualified as safe, risky, or important. Furthermore, a non-home loca-tion, such as a friend’s house, can be identified as a ‘‘risky’’ place during thebase-line interview but later rated as a ‘‘safe’’ place. Our preliminary GIS ana-lysis narrows our sample down to 56 subjects as 2 subjects do not haverecords associated with the base-line interview. Of these 56 subjects, 20had 3 month follow-ups, 10 had 6 month follow-ups, and only 2 had 12month follow-ups.
During the base-line interviews, the home addresses for all 56 subjectswere collected. These exact locations were kept in a database with strictconfidentiality. Utilizing GIS, we geocoded these addresses using the U.S.Census Bureau’s TIGERs line files of street networks of eight jurisdictions (Dis-trict of Columbia; Montgomery County and Prince George’s County ofMaryland; Arlington County, Fairfax County, Alexandria City, Falls ChurchCity, and Fairfax City of Virginia). Four locations were excluded due to incom-plete geographic data, or because the subject’s home address was outside ofour catchment area of the DC metropolitan region. The majority of our subjects
24 M. Mason et al.
Dow
nloa
ded
by [
Am
s/G
iron
a*ba
rri L
ib]
at 0
3:27
15
Oct
ober
201
4
were located within the Capitol Beltway, the District of Columbia, or live in theinner-suburbs, with only 14 subjects located outside the Beltway.
For this article, we focused on the 23 subjects from the District ofColumbia in order to illustrate our method with an urban sample. Weobtained 111 locations from the interviews from this narrowed sample.Sixty-seven locations were not home locations. We termed these non-homelocations ‘‘activity locations.’’ For example, these locations can be cinemas,fast food restaurants, parks, metro stops, and homes of friends. In otherwords, these are simply locations where our subjects ‘‘hang-out’’ and areactive with their social networks. Spatially, 42 of these activity locations arelocated in DC (63%), 19 in Maryland (28%), and 6 in Virginia (9%). Weattempted to geocode all 42 DC activity locations. Unfortunately, these loca-tions were not reported uniformly with high address accuracy. Locationswere reported by our subjects in four ways. First, the most desirable methodis street address. Second, although less desirable, the use of intersection oftwo streets is not uncommon; for example, a teen may say ‘‘it’s near NorthCapital and M street’’ (Levine & Kim, 1998). Most GIS software with geo-coding function is capable of handling this type of location reference. How-ever, success rate is considerably lower than using street addresses. Third,our subjects reported activity locations using business name (e.g., UptownTheater) or place name (e.g., near Dupont Circle). These references requiredmanual geocoding that was very time consuming and sometimes imprecise.Fourth, sometimes there was no reference given by our subject. For example,an activity location may simply be ‘‘the lot on Florida avenue.’’ Even afterprompting the teen for cross-roads, retail outlets, landmarks, and using amap, some were not be able to identify the specific geographic location.We were able to successfully geocode 36 (54%) DC activity locations, andthese formed the basis for subsequent analyses.
Protective Location Results
We created an initial geospatial database that used nine commonly acceptedprotective and risky categories as an initial guide toward understanding thesocial ecology of urban youth: Protective locations: (1) health services{measured from home location}; (2) recreational facilities {measured fromhome location}; (3) social services {measured from home location}; (4) reli-gious resources {measured from home and activity locations}; (5) police sta-tions {measured from home and activity locations}; (6) entertainmentopportunities {measured from home and activity locations}. Risk locations:(7) alcohol outlets {measured from home and activity locations}; (8) crime inci-dences {measured from home and activity locations}; (9) deteriorated housing{measured from home and activity locations} (Mason et al., 2005; Macintyre &Ellaway, 2003; Raduenbus, 2003). We anticipate our categories to evolve indetail and type as well as in quantity of resources as we learn more from the
Creating a Geospatial Database 25
Dow
nloa
ded
by [
Am
s/G
iron
a*ba
rri L
ib]
at 0
3:27
15
Oct
ober
201
4
teens. So far we have four protective categories geocoded: public libraries,Boys and Girls Clubs, public parks, and police stations.
Many metropolitan areas have government-sponsored GIS data avail-able to the public. In DC we accessed the Office of the Chief TechnologyOfficer, which provided myriad GIS data layers through the DC GIS onlineportal (see http://dcgis.dc.gov). These data layers were in shapefile format-ting. We obtained LIBRARY, PARK, and POLICE data layers from DC GIS.We obtained a list of all BGCLUBs and then geocoded all of these addresses.All four data layers are constructed using the same map projection and coor-dinate system using ESRI ArcGIS (ArcView). In most GIS software data isorganized in themes as data layers. This approach allows data to be inputas separate themes and overlaid based on analysis requirements. This canbe conceptualized as vertical layering the characteristics of the earth’s sur-face. Figure 1 shows 52 home locations. The map was deliberately producedwith poor spatial accuracy to provide privacy. Further, the actual locations ofall participants’ home and other identified places were randomly relocated
FIGURE 1 Geography of positive resources and the subjects’ home and activity locations.
26 M. Mason et al.
Dow
nloa
ded
by [
Am
s/G
iron
a*ba
rri L
ib]
at 0
3:27
15
Oct
ober
201
4
within the respective one square kilometer area, thereby eliminating thechance of connecting geographic data to participant identities. Figure 1 alsoshows all four types of positive resources overlaid with both home and activ-ity locations (with self-reported ‘‘safety’’ rating) obtained from the interviews.The positive resources, when taken together, appeared to be evenly distrib-uted across DC. The visualization did not reveal any discernable pattern.
In order to establish the spatial relationship between the positiveresources and the activity locations, we used ArcView Network Analyst toconstruct travel bands (a representation of street networks from an indexpoint {e.g., home location} to a given parametric distance {e.g., 1 kilometer})around libraries, Boys & Girls Clubs, and police stations. Network Analystallowed us to follow the actual street networks when connecting two loca-tions, allowing for more accurate travel analyses between two locations.Building on these distance assignments, we constructed travel bands using1-kilometer increments up to the maximum of 10 kilometers. The same pro-cesses are simultaneously done for all six Boys & Girls Clubs. Then we simi-larly constructed travel bands with the same increment and maximumdistance for libraries and police stations. Figure 2 shows the two kilometer(1.25 miles) travel bands from all 6 Boys & Girls Clubs (left panel) and all16 police stations and sub-stations. It is important to note that Figure 3 showsonly one travel band. In reality there were 10 travel bands for each location.An interesting observation is the lack of Boys & Girls Clubs coverage inthroughout NW DC. We located only six Boys & Girls Clubs in the entire
FIGURE 2 Two-kilometer travel bands constructed from (a) Boys & Girl Clubs and (b) policestations.
Creating a Geospatial Database 27
Dow
nloa
ded
by [
Am
s/G
iron
a*ba
rri L
ib]
at 0
3:27
15
Oct
ober
201
4
District and five of these are in the SW and SE quadrants of the city. Also,even with 2-kilometer travel bands, there were considerable overlaps ofpolice station proximity, especially in areas immediately north and east ofdowntown, in Northwest DC, giving the impression of clusters of policepresence in some areas and little to no coverage in other areas.
With the travel bands constructed around the positive resources, weperformed a process known as ‘‘spatial-join’’ to assign travel distancebetween each of the 36 activity locations to each of the positive resourcelocations. Spatial-join is a type of table-join operation in which fields fromone layer’s attribute table are appended to another layer’s attribute tablebased on the relative locations of the features in the two layers. As a demon-stration we used BGCLUB and POLICE. Because there are 6 BGCLUBs and 16police stations, we constructed a matrix to capture the travel distance rela-tionship. First, we created a 36 (activity locations) by 6 (Boys & Girls Clubs)distance matrix (Table 1) that included distance values and self-reported‘‘importance’’ and ‘‘safety’’ and ‘‘risk’’ ratings. The last six columns displaythe travel distance (in meters) between the activity locations and the clubs.Blank entry represents distance beyond 10 kilometers. Similarly, a 36 by 16distance matrix was constructed to capture the travel distance betweenactivity locations and police stations.
Risk Location Results
We used two commonly referenced neighborhood-level features to representrisk resources. Robbery incidences, like other crimes, are often used to study
FIGURE 3 Activity locations and density of (a) robbery incidence and (b) liquor stores.
28 M. Mason et al.
Dow
nloa
ded
by [
Am
s/G
iron
a*ba
rri L
ib]
at 0
3:27
15
Oct
ober
201
4
the quality of a neighborhood. We speculated that high incidence of rob-beries may be viewed as a form of neighborhood disorder that may translateinto a fear of crime by our subjects whose activity locations are situated inthese areas. Similarly, the presence of alcohol outlets, such as liquor stores,may also have been perceived by our subjects as a sign of ‘‘danger’’ as theliterature has long linked alcohol consumption and human violence (Kumar& Waylor, 2003).
We obtained over 3,500 listings of robbery incidences in 2002 from theonline edition of the Washington Post (see www.washingtonpost.com).These listings were abbreviated versions of the police reports obtained fromthe DC Metropolitan Police. The listings showed the locations of robbery
Table 1 Travel Distance Between DC Activity Locations and Boys & Girls Clubs
SUBJECT#Importance
(Self)Safety
Rating (Self) BG01 BG02 BG03 BG04 BG05 BG06
2700010002 unrated 1000 6000 5000 8000 100002700010003 yes rated risky 2000 7000 6000 90002700010005 yes unrated 2000 7000 6000 90002700010006 no rated risky 2000 7000 6000 90002700010007 no rated safe 8000 4000 4000 6000 5000 90002700040003 yes unrated 7000 7000 3000 70002700040004 no rated risky 50002700040006 yes rated safe 7000 7000 3000 70002700060002 yes rated both 8000 8000 5000 70002700060003 no rated safe 8000 8000 5000 70002700060004 no rated safe 7000 8000 9000 6000 80002700070002 yes unrated 3000 5000 3000 7000 80002700070003 no rated safe 3000 5000 3000 7000 80002700110002 yes unrated 8000 8000 2000 90002700110003 no rated risky 7000 8000 6000 7000 100002700140003 no rated risky 50002700140006 no rated risky 7000 8000 6000 7000 100002700140008 yes rated safe2700190002 yes rated risky 8000 8000 5000 70002700220003 no rated safe 6000 7000 10000 5000 70002700230002 yes rated both 8000 3000 4000 8000 2000 60002700240002 no rated risky 4000 3000 2000 10000 5000 80002700250002 no rated risky 9000 5000 5000 5000 5000 90002700250003 yes rated safe 9000 5000 5000 5000 5000 90002700300003 no rated risky 40002700300006 no rated risky 50002700300008 no rated risky 8000 8000 4000 70002700300010 yes unrated 50002700350002 yes rated safe 4000 4000 3000 6000 80002700500002 no rated risky 60002700500003 yes rated safe 70002700500005 no rated safe 8000 8000 2000 90002700500006 yes rated risky 10000 10000 5000 100002700730003 yes unrated 40002700730006 yes unrated 40002700930003 no rated risky 8000 4000 4000 7000 3000 7000
Creating a Geospatial Database 29
Dow
nloa
ded
by [
Am
s/G
iron
a*ba
rri L
ib]
at 0
3:27
15
Oct
ober
201
4
incidences. Due to the large number of incidences, we decided to construct adensity surface map (a visual display of the number of incidences per squareunit) to show the uneven distribution of robbery incidences in DC. Wespeculated that, like other property crimes such as car theft, robbery perpe-trators were likely to live in the general vicinity of the robbed home (Levineet al., 2000; Lu, 2003). Similarly, we speculated that robbery incidences arelargely a local occurrence. Even though our subjects are fairly mobile, weexpected them to interact, at some level, that is, spending some time athome, with their local neighborhood from their activity locations. Therefore,when constructing our density surface, we chose to use one kilometer as the‘‘search distance’’ to search for robbery incidences. Using ArcGIS SpatialAnalyst, we chose the Kernel Density Estimation approach to construct thedensity surface. This estimation approach accounts for the influence of dis-tance (distance-decay mechanism). In other words, robbery incidences wereexpected to have stronger (more negative) influence to areas closer thanthose areas farther away.
We obtained the addresses of liquor stores in DC from the local phonedirectories. Using the same method as explained earlier, we generated den-sity surface for liquor stores. The density surfaces are shown in Figure 3. MapA (left) clearly depicts a ‘‘hot-spot’’ of high robbery incidences in 2002. Map B(right) shows a higher concentration of liquor stores slightly south of thebright red robbery hotspot. In both figures, the activity locations self-ratedas ‘‘risky’’ are also displayed. In general, some degree of spatial correlationbetween these ‘‘risky’’ activity locations and the clusters of high robbery inci-dences and liquor stores are apparent.
Self-Reported Safety Level by Protective and Risky Location Results
We further analyzed Table 1 data by extracting the travel distances (inmeters) between 36 activity locations and police stations and Boys & GirlsClubs. Using five kilometers as the threshold, we tallied the total numbersof police stations and Boys & Girls Clubs, respectively, for each of the 36activity locations. Five kilometers represented an average travel band. Bothapproaches estimated the geographical accessibility of the two protectiveresources, thereby producing a protection exposure estimate. We selectedsix parameters for this preliminary analysis: (1) estimated density of liquorstores; (2) estimated density of robbery incidences; (3) travel distance to clo-sest Boys & Girls Clubs; (4) number of Boys & Girls Clubs within five kilo-meters of travel distance; (5) travel distance to closest police stations; and(6) number of police stations within five kilometers of travel distance. Asan initial analytic step with this database, we tested for significant differencesbetween the travel distances from the respondents’ homes to (i) the closestBoys & Girls Club, (ii) the activity locations described as safe, and (iii) theactivity locations described as risky in order to examine the relationship
30 M. Mason et al.
Dow
nloa
ded
by [
Am
s/G
iron
a*ba
rri L
ib]
at 0
3:27
15
Oct
ober
201
4
between these variables. Our early work had shown geographical variancethat was statistically significant (Mason et al., 2004). The ensuing t-testsrevealed no statistically significant differences.
DISCUSSION
Limitations
We acknowledge that there are limitations of our preliminary analysis of the23 DC cases. First, we have limited our activity locations to within DC; incontrast there are many activity locations in Maryland. By adding those loca-tions in Maryland, we should be able to establish a more comprehensiverepresentation of our DC subjects’ activity locations. Second, as discussedearlier, we have only demonstrated two simplistic approaches to characterizegeographical access to protective resources (i.e., distance of nearestresources and number of resources within a threshold distance). Third, eventhough our travel band analyses provide a far more realistic estimate of traveldistance than using Euclidean (straight line) distance, we have not adoptedthe use of travel time by addressing mode of travel. Fourth, as discussedby Kumar and Waylor (2003), proximity of alchohol-serving establishmentsand crime probabilities are linked. This shows a very high degree of colli-nearity between our parameters. We have found that generalizing even‘‘known’’ risk locations such as liquor stores may be confounded by adoles-cents’ interpretation of high-crime locations as being safe, for example.Therefore, we acknowledge the iterative nature of our work in order tounderstand modern urban youth.
Fifth, we have only explored very limited protective resources. Thereare many other resources that may influence the sense of safety for our sub-jects and healthful outcomes. Last, but not least, sometimes our subjects self-report a location as ‘‘safe’’ during one phase of the interview and later regardthe same location as ‘‘risky’’ during a follow-up interview. This may incorpo-rate error into our preliminary analysis, or it could be that the locations havechanged from ‘‘safe’’ to ‘‘risky,’’ or it may be a feature of urban teens’ fluidperceptions and interpretations of their spatial worlds.
CONCLUSIONS
Despite the challenges of using GIS for the study of urban youth, we believethat GIS is an essential tool in aiding our construction of a geography ofactivity locations in order to more fully understand the unique social ecologyof urban youth. Our work fills a need by describing micro-level geographicdata that influence urban adolescent health behavior, including contextualand sociospatial variables. This methodology produces highly relevant
Creating a Geospatial Database 31
Dow
nloa
ded
by [
Am
s/G
iron
a*ba
rri L
ib]
at 0
3:27
15
Oct
ober
201
4
prevention data that can guide targeted Universal, Selective, and Indicatedlevel interventions. The next step in the application of this technology willbe to streamline the data collection, analyses, and presentation of thegeographic data in order to apply these to real-life settings, such as primarycare and community centers, and urban schools. For example, our work iscurrently focused on strategies to apply this technology within the uniquesetting of adolescent primary care clinics (Walker, Mason, & Cheung,2006). Specific, empirically derived social-ecological profiles, can be usedas a launching point for individualized, relevant interventions to teach youthprevention skills that are environmentally based. For example, we anticipateour study will produce data that can inform interventions with teens as tohow to utilize social networks to increase time spent in safe locations withsafe network members, and the like. Such personalized interventions willbe more relevant, highly specific, and thus likely to be more effective.
GLOSSARY
Adjacency: A spatial concept describing the state of being adjacent; contiguity.Connectivity: A spatial concept describing spaces that are serving or tending
to connect.Density Surface Map: A visual display of the number of incidences per square
unit with the color spectrum representing density variation.Geocoding=Georeferencing: A process by which a specific location is given a
label identifying its location with respect to some common referencepoint.
Travel Bands: A representation of street networks from a geographic indexpoint to a given parametric distance.
NOTE
1. For details on the social network measure, see Mason et al. (2004).
REFERENCES
Berkman, L., & Clark, C. (2003). Neighborhoods and networks: The construction ofsafe places and bridges. In I. Kawachi & L. Berman (Eds.), Neighborhoods andhealth (pp. 288–302). New York: Oxford University Press.
Burrough, P. (1986). Principles of geographical information systems for landresource assessment. Oxford: Clarendon Press.
Cheadle, A., Wagner, E., Koepsell, T., Kristal, A., & Patrick, D. (1992). Environmentalindicators: A tool for evaluating community-based health-promotion programs.American Journal of Preventive Medicine, 8, 345–350.
32 M. Mason et al.
Dow
nloa
ded
by [
Am
s/G
iron
a*ba
rri L
ib]
at 0
3:27
15
Oct
ober
201
4
Clarke, K. C. (2003) Getting started with geographic information systems (4th ed.).Upper Saddle River, NJ: Prentice Hall.
Crouter, A., & Larson, R. (1998). Editors notes. In A. Crouter & R. Larson (Eds.),Temporal rhythms in adolescence: Clocks, calendars, and the coordination ofdaily life (pp. 1–6). San Francisco: Jossey-Bass.
Duncan, G., & Aber, L. (1997). Neighborhood models and measures. In J.Brooks-Gunn, G. J. Duncan, & L. J. Aber (Eds.), Neighborhood poverty: Policyimplication in studying neighborhoods (pp. 62–78). New York: Russell SageFoundation.
Furstenberg, F., & Hughes, M. (1997). The influence of neighborhoods on children’sdevelopment: A theoretical perspective and a research agenda. In J.Brooks-Gunn, G. J. Duncan, & L. J. Aber (Eds.), Neighborhood poverty: Policyimplication in studying neighborhoods (pp. 23–47). New York: Russel SageFoundation.
Kearns, R., & Moon, G. (2002). From medical to health geography: Novelty, placeand theory after a decade of change. Progress in Human Geography, 26(5),605–625.
Kumar, N., & Waylor, C. R. M. (2003). Proximity to alcohol-serving establishmentsand crime probabilities in Savannah, Georgia: A statistical and GIS analysis.Southeastern Geographer, 43(1), 125–142.
Larson, R. (1998). Implications for policy and practice: Getting adolescents, families,and communities in sync. In A. Crouter & R. Larson (Eds.), Temporal rhythms inadolescence: Clocks, calendars, and the coordination of daily life (pp. 37–52).San Francisco: Jossey-Bass.
Levine, N., & Associates (2000). CrimeStat: A spatial statistics program for the ana-lysis of crime incident locations (version 1.1). Annandale, VA: Ned Levine andAssociates; Washington, DC: The National Institute of Justice.
Levine, N., & Kim, K. E. (1998). The location of motor vehicle crashes in Honolulu: Amethodology for geocoding intersections. Computers, Environment and UrbanSystems, 22(6), 557–576.
Lu, Y. 2003. Getting away with the stolen vehicle: An investigation of journey-after-crime. Professional Geographer, 55(4), 422–433.
Macintyre, S., & Ellaway, A. (2003). Neighborhoods and health: An overview. InL. Berman, Neighborhoods and health (pp. 20–42). New York: Oxford Univer-sity Press.
Mason, M., Cheung, I., & Walker, L. (2004). Substance use, social networks andthe geography of urban adolescents. Substance Use and Misuse, 39(10–12),1751–1778.
Raudenbush, S. (2003). The quantitative assessment of neighborhood social environ-ments. In I. Kawachi & L. Berman (Eds.), Neighborhoods and health (pp. 112–131). New York: Oxford University Press.
Sampson, R. J. (2003). Neighborhood-level context and health: Lessons from sociol-ogy. In L. Berman (Ed.), Neighborhoods and health (pp. 132–146). New York:Oxford University Press.
Sack, R. (1997). Homo geographicus. Baltimore: Johns Hopkins University Press.Star, J., & Estes, J. (1990). Geographic information systems: An introduction. Engle-
wood Cliffs, NJ: Prentice Hall.
Creating a Geospatial Database 33
Dow
nloa
ded
by [
Am
s/G
iron
a*ba
rri L
ib]
at 0
3:27
15
Oct
ober
201
4
Sussman, S., Ames, S., Dent, C., & Stacy, A. (2001). Self-reported high-risk locationsof drug use among drug offenders. American Journal of Drug and AlcoholAbuse, 27(2), 281–299.
Walker, L., Mason, M., & Cheung, I. (2006). Adolescent substance use and abuse pre-vention and treatment: Primary care strategies involving social networks and thegeography of risk and protection. Journal of Clinical Psychology in MedicalSettings. 13(2), 126–134.
Weller, S., & Romney, K. (1998). Systematic data collection. Thousand Oaks, CA:Sage.
34 M. Mason et al.
Dow
nloa
ded
by [
Am
s/G
iron
a*ba
rri L
ib]
at 0
3:27
15
Oct
ober
201
4
